Numerators and denominators define what a percentage in a nursing dissertation actually means. The numerator counts the event or characteristic of interest. The denominator defines the eligible whole to which that count refers. If the two do not share compatible people, units, periods and rules, a precise-looking percentage can answer the wrong question.
This guide examines which records belong on each side of a service-evaluation proportion. Its practical output is a counting specification and reconciliation check for eligible episodes, qualifying events and unknown status. Begin with the how to choose indicators for a nursing service evaluation if you first need to establish the broader project approach. Use the detailed decisions below once that foundation is clear.
This guide concentrates on constructing and checking service-evaluation proportions. It does not repeat the broader selection of quality indicators, and it does not replace specialist analysis of event rates or person-time. The worked numbers are fictional teaching examples. Their purpose is to make counting decisions visible before a student interprets service performance.
Start with the question that the count should answer
Write the question in ordinary language before choosing a formula. “How many eligible follow-up episodes had a documented plan?” concerns episodes. “How many patients received a follow-up contact?” concerns distinct people. “How many contacts did the team complete?” concerns activity. These are related service questions, but their numerators and denominators are different.
Then state what the resulting number would and would not establish. A documented plan can describe recording completeness. It does not prove that a patient understood the plan or received every intended action. The arithmetic should remain connected to the precise service issue, rather than lending a broader quality claim to whichever data happen to be available.
Define numerators and denominators in a measurement sentence
Use a full sentence: “The numerator is the number of eligible episodes with a plan recorded within the defined window; the denominator is all eligible episodes entering that window during the period.” Replace every vague word with an operational definition appropriate to the project. A reader should understand which records qualify without asking you to reconstruct the rule later.
For a proportion, the numerator is a subset of the denominator. If a record can enter the numerator without qualifying for the denominator, inspect the definitions. For example, plans from episodes outside the reporting period should not be counted against episodes inside it. The calculation may otherwise exceed its intended meaning even if the spreadsheet formula is correct.
Distinguish people, episodes, opportunities and contacts
One person may have several episodes, and one episode may contain several contacts. An opportunity can refer to a defined occasion on which an action should be considered or recorded. Decide which unit addresses the question and retain that unit throughout the methods, results and discussion.
Repeated records are not necessarily errors. They may be legitimate contacts when measuring workload. They become a problem when counted as distinct people in a patient-coverage proportion. Use appropriate identifiers and a documented deduplication rule within the project arrangements. Do not merge records on a weak assumption merely because two entries look similar.
Specify eligibility before examining the result
Define the service, episode type, relevant period and any justified exclusions before calculating the percentage. A denominator is not simply everything that remains after inconvenient records have been removed. Each boundary needs a reason tied to the question and the available evidence.
If eligibility is uncertain, distinguish “not eligible” from “eligibility not established”. Missing information does not automatically justify exclusion. Explain how unresolved records are handled and how that choice affects the interpretation. The separate guide on inclusion and exclusion criteria addresses the wording and application of eligibility rules in more detail.
Worked example: documented plans in eligible episodes
Imagine a fictional evaluation with 80 eligible episodes during a defined reporting period. A plan meeting the operational recording rule is present in 60. The proportion is 60 divided by 80, multiplied by 100, which equals 75%. Report the counts beside the percentage: 60 of 80 eligible episodes, or 75%.
Now suppose a student divides the 60 plans by 100 total contacts because that number is easier to obtain. The result is 60%, but it does not represent the same question. The numerator concerns episodes with plans; the denominator concerns contacts. The difference is a definition error rather than evidence that performance changed.
The appropriate results sentence remains bounded: “In this fictional dataset, a qualifying plan was recorded for 60 of 80 eligible episodes.” It should not say that 75% of patients received high-quality care. The measurement concerns a defined record within an episode, and broader claims would need different evidence.
Use a counting-rule table
| Decision | Illustrative rule | Check |
|---|---|---|
| Unit | One eligible episode | Repeated contacts do not create extra episodes |
| Numerator | Episode with a qualifying recorded plan | Every counted episode is in the denominator |
| Denominator | All eligible episodes within the stated window | Exclusions have a documented rationale |
| Unknown status | Recording status cannot be established | Reported separately from confirmed absence |
| Period | Defined entry and observation dates | The numerator and denominator use compatible boundaries |
This original table is a planning aid. It does not define a clinical standard or establish that any particular recording rule is appropriate for your service. Confirm the relevant local definitions before applying it.
Handle missing outcomes without changing the question silently
Suppose the fictional 80 episodes include ten whose plan status cannot be established from the available extract. You need to describe what is known. A confirmed absent plan and an unknown recording status are different categories. Whether the indicator’s agreed rules treat unavailable documentation as a failure is a separate question that must be explicit.
One report could show the numbers with a qualifying plan, without one and with unknown status. If presenting a proportion among records with known status, label that denominator clearly and retain the missing count. Do not quietly remove unknown cases and describe the resulting percentage as covering all eligible episodes.
Keep nonresponse separate from service non-delivery
A patient questionnaire may ask whether a follow-up contact occurred. People who do not respond have an unknown answer, not automatically an absent contact. The response denominator and the service-delivery denominator answer different questions. State which one the calculation uses.
For example, 30 of 40 respondents reporting a contact means 75% of respondents reported that experience. If 100 eligible people were invited, it does not establish that 75% of all invited people received the contact. The views and experiences of nonrespondents remain uncertain. Avoid extending the denominator in the prose after narrowing it in the calculation.
Check the time boundary and the opportunity to qualify
An episode entering near the end of a reporting period may not yet have had the full opportunity to meet a follow-up rule. Treating it as a completed failure can distort the indicator. Define the entry event, allowable observation window and extraction date before selecting the records.
This is not a reason to invent a universal waiting period. Use the question and relevant service definition to establish the window. If the project cannot observe complete episodes, state the limitation or narrow the population appropriately. The separate timeframe discussion concerns this design choice; the calculation here depends on applying the chosen boundary consistently.
Distinguish a proportion from an event rate
A proportion describes a subset of a defined whole. An event rate can relate repeated events to an exposure such as time at risk. Those measures require different interpretations. A person can experience several events, so an event numerator need not be a subset of a count of people in the same way.
Do not label every division a percentage. State the unit in the result, including any multiplier. If your project concerns events per patient days, use the dedicated patient-day rate guide for the relevant distinction. This article instead addresses denominator construction for bounded service-evaluation proportions.
Pool periods by adding compatible counts
Consider a fictional first period with 10 qualifying episodes out of 20 and a second with 45 out of 50. Their percentages are 50% and 90%. The combined proportion is 55 divided by 70, approximately 78.6%. Averaging 50% and 90% gives 70%, which gives the two periods equal weight despite their different denominator sizes.
Only pool counts when the definitions are compatible. A changed eligibility rule, altered recording window or different service population may make a combined figure misleading. Explain any such change before deciding whether aggregation is useful. Correct arithmetic cannot make incompatible definitions comparable.
Do not confuse percentage points with percentage change
In a fictional comparison, an indicator rises from 60% to 75%. The absolute increase is 15 percentage points. The increase relative to the original percentage is 15 divided by 60, or 25%. These statements describe the same numerical movement in different ways.
Use the wording that suits the question and report the underlying counts. Neither form establishes that a service change caused the difference. Small denominators, different populations and recording changes may affect the comparison. Be especially careful when a dramatic relative figure disguises a small number of records.
Report zero, unavailable and not applicable distinctly
A numerator of zero with a valid positive denominator can be a real result. A denominator of zero means that the proposed proportion cannot be calculated for that period. Missing denominator information is another condition again. None should be converted automatically into 0% merely to keep a chart complete.
Choose clear labels and explain them in the table notes. “No eligible episodes”, “denominator unavailable” and “zero qualifying episodes” convey different information. A reader should not have to infer which condition applies from a blank cell or a line drawn down to zero.
Check the data path from record to result
Trace a small set of appropriately handled records through eligibility, classification and counting. Confirm that the spreadsheet includes each eligible unit once and that the numerator rule has been applied consistently. Keep a decision log for ambiguous examples. This is a verification exercise, not permission to access additional patient information.
If records come from more than one system, document the linkage or reconciliation process. Do not assume identical labels mean identical units. A completed appointment in one system may not equal a completed episode in another. Record the source and definition beside the final table so the calculation remains reviewable.
Present numerators and denominators with their interpretation limits
NHS Institute measurement guidance includes ratios, percentages and a measures template (NHS Institute for Innovation and Improvement, n.d.). Use the general principle of explicit measurement definitions, while writing your own service-specific rules. A template supports transparency; it does not determine the correct denominator for every project.
A results table should show the period, qualifying count, eligible count, percentage where meaningful and relevant unknowns. Add a short note on the unit and exclusions. The discussion should then explain what the indicator says about the chosen process, without extending a record-based result into an unsupported claim about patient benefit.
Practical denominator audit
- Write the question in words and identify the counting unit.
- Define eligibility and the observation window.
- Confirm that every numerator record belongs to the denominator.
- Distinguish repeated contacts from distinct people or episodes.
- Report unknown status and nonresponse explicitly.
- Check zero denominators and unavailable information.
- Pool compatible counts rather than averaging unlike percentages.
- Retain counts, units and interpretation boundaries in the results.
Ask another reader to apply the rules to a few fictional borderline records. Differences in their decisions can reveal ambiguous wording before the analysis is finalised. Resolve the rule through the appropriate project process and document any change.
Label the final figure precisely
Use a label that names the counted unit, eligibility and recorded event. “Recorded plans among eligible episodes” is more informative than “success rate”. The wording should remain consistent in the table, abstract and discussion so that the same number does not acquire a broader meaning as the dissertation progresses.
Frequently asked questions
Can I use all available records as the denominator?
Only if they represent the eligible whole required by the question. Availability and eligibility are different. Explain the frame, period and exclusions, and identify which intended records may be missing rather than letting convenience determine the measure.
Should missing records count as failures?
That depends on the explicitly agreed indicator definition and what the missing information means. Confirmed absence, unavailable documentation and an unobserved outcome should not be conflated without explanation. Report uncertainty and avoid changing the rule after seeing the result.
Can I average monthly percentages?
A combined proportion generally requires compatible numerator and denominator counts. Adding those counts gives periods their appropriate contribution. An unweighted average answers a different question and can mislead when period sizes differ substantially.
What if the denominator is zero?
The proportion is not calculable for that period. Label the reason rather than reporting 0%. Distinguish a genuine absence of eligible units from unavailable denominator information, because the two conditions have different implications.
Does a higher percentage prove better nursing care?
It may show a change in the defined indicator. It does not establish overall care quality or causation. Explain the measure’s meaning, data quality, population and period before discussing its relevance to a nursing service.
Resolve duplicate records before counting eligible episodes
Duplicate records create a different problem from repeated care. An episode may appear twice because two exports overlap, because a form was resubmitted, or because a referral generated an administrative copy. Counting both records would enlarge the denominator without representing an additional eligible episode. However, two genuine visits by the same patient should remain separate when visits are the chosen counting unit.
Write a deduplication rule before looking at whether a record contributes to the numerator. For example, an approved episode identifier and service date might establish that two rows represent one event. The rule should specify which source takes priority when duplicated rows contain conflicting completion fields. Do not resolve a conflict by choosing whichever entry improves the percentage.
Keep a restricted processing log that records the number of duplicates removed and the reason for removal. The dissertation normally needs the aggregate decision trail, not identifiable source rows. If reliable linkage is unavailable, describe the remaining uncertainty instead of presenting a guessed count as exact. An unresolved duplication problem can affect both sides of the fraction, so a tidy percentage does not establish a clean underlying dataset.
Use a reconciliation check before presenting the result
A practical final check asks whether every eligible episode has exactly one outcome category. In the fictional example, the eighty eligible episodes might comprise sixty confirmed plans, twelve confirmed absences and eight records with unknown status. Those categories sum to eighty. If they sum to eighty-three, investigate overlapping categories or duplicate episodes before interpreting the result.
Ask someone authorised to review the method to trace a small approved sample from source categories into the final count. The purpose is to check the rule, not to disclose records to an external writer or to obtain a second clinical judgement. Discuss any disagreement about coding and record the decision consistently across the dataset. A reproducible calculation needs both a clear formula and a clear route from records to that formula.
Connect the calculation to your dissertation question
Read the parent guide on choosing service-evaluation indicators before deciding what to count. Use the population and sampling guide if coverage is the main uncertainty. For feedback on an anonymised measurement plan, explore our nursing evaluation support or send a focused enquiry. Educational support should help you justify the calculation, not replace your own analysis.
References
NHS Institute for Innovation and Improvement (n.d.) The how-to guide for measurement for improvement. Available at: NHS England (Accessed: 16 October 2026).
